Enhancing LLM Watermark Resilience against Both Scrubbing and Spoofing Attacks

Watermarking is a promising defense against the misuse of large language models LLMs, yet it remains vulnerable to scrubbing and spoofing attacks. This vulnerability stems from an inherent trade-off governed by watermark window size: smaller windows resist scrubbing better but are easier to...

BlowPrint: Blow-Based Multi-Factor Biometrics for Smartphone User Authentication

Biometric authentication is a widely used security mechanism that leverages unique physiological or behavioral characteristics to authenticate users. In multi-factor biometrics MFB, multiple biometric modalities, e.g., physiological and behavioral, are integrated to mitigate the limitations...